1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
|
import operator_benchmark as op_bench
import torch
import torch.nn as nn
"""
Microbenchmarks for MaxPool1d and AvgPool1d operators.
"""
# Configs for pool-1d ops
pool_1d_configs_short = op_bench.config_list(
attr_names=[
'kernel', 'stride', 'N', 'C', 'L'
],
attrs=[
[3, 1, 8, 256, 256],
],
cross_product_configs={
'device': ['cpu', 'cuda'],
},
tags=['short']
)
pool_1d_configs_long = op_bench.cross_product_configs(
kernel=[3],
stride=[1, 2],
N=[8, 16],
C=[3],
L=[128, 256],
device=['cpu', 'cuda'],
tags=['long']
)
pool_1d_ops_list = op_bench.op_list(
attr_names=['op_name', 'op_func'],
attrs=[
['MaxPool1d', nn.MaxPool1d],
['AvgPool1d', nn.AvgPool1d],
],
)
class Pool1dBenchmark(op_bench.TorchBenchmarkBase):
def init(self, kernel, stride, N, C, L, device, op_func):
self.inputs = {
"input": torch.rand(N, C, L, device=device)
}
self.op_func = op_func(kernel, stride=stride)
def forward(self, input):
return self.op_func(input)
op_bench.generate_pt_tests_from_op_list(pool_1d_ops_list,
pool_1d_configs_short + pool_1d_configs_long,
Pool1dBenchmark)
"""
Microbenchmarks for MaxPool2d and AvgPool2d operators.
"""
# Configs for pool-2d ops
pool_2d_configs_short = op_bench.config_list(
attr_names=[
'kernel', 'stride', 'N', 'C', 'H', 'W'
],
attrs=[
[[3, 1], [2, 1], 1, 16, 32, 32],
],
cross_product_configs={
'device': ['cpu', 'cuda'],
},
tags=['short']
)
pool_2d_configs_long = op_bench.cross_product_configs(
kernel=[[3, 2], [3, 3]],
stride=[[2, 2]],
N=[8, 16],
C=[32],
H=[32, 64],
W=[32, 64],
device=['cpu', 'cuda'],
tags=['long']
)
pool_2d_ops_list = op_bench.op_list(
attr_names=['op_name', 'op_func'],
attrs=[
['MaxPool2d', nn.MaxPool2d],
['AvgPool2d', nn.AvgPool2d],
['AdaptiveMaxPool2d', lambda kernel, stride: nn.AdaptiveMaxPool2d(kernel)],
['FractionalMaxPool2d', lambda kernel, stride: nn.FractionalMaxPool2d(kernel, output_size=2)],
],
)
class Pool2dBenchmark(op_bench.TorchBenchmarkBase):
def init(self, kernel, stride, N, C, H, W, device, op_func):
self.inputs = {
"input": torch.rand(N, C, H, W, device=device)
}
self.op_func = op_func(kernel, stride=stride)
def forward(self, input):
return self.op_func(input)
op_bench.generate_pt_tests_from_op_list(pool_2d_ops_list,
pool_2d_configs_short + pool_2d_configs_long,
Pool2dBenchmark)
"""
Microbenchmarks for MaxPool3d and AvgPool3d operators.
"""
# Configs for pool-3d ops
pool_3d_configs_short = op_bench.config_list(
attr_names=[
'kernel', 'stride', 'N', 'C', 'D', 'H', 'W'
],
attrs=[
[[3, 1, 3], [2, 1, 2], 1, 16, 16, 32, 32],
],
cross_product_configs={
'device': ['cpu', 'cuda'],
},
tags=['short']
)
pool_3d_configs_long = op_bench.cross_product_configs(
kernel=[[3, 2, 3], [3, 3, 3]],
stride=[[2, 2, 2]],
N=[8, 16],
C=[32],
D=[32],
H=[32, 64],
W=[32, 64],
device=['cpu', 'cuda'],
tags=['long']
)
pool_3d_ops_list = op_bench.op_list(
attr_names=['op_name', 'op_func'],
attrs=[
['MaxPool3d', nn.MaxPool3d],
['AvgPool3d', nn.AvgPool3d],
['AdaptiveMaxPool3d', lambda kernel, stride: nn.AdaptiveMaxPool3d(kernel)],
['FractionalMaxPool3d', lambda kernel, stride: nn.FractionalMaxPool3d(kernel, output_size=2)],
],
)
class Pool3dBenchmark(op_bench.TorchBenchmarkBase):
def init(self, kernel, stride, N, C, D, H, W, device, op_func):
self.inputs = {
"input": torch.rand(N, C, D, H, W, device=device)
}
self.op_func = op_func(kernel, stride=stride)
def forward(self, input):
return self.op_func(input)
op_bench.generate_pt_tests_from_op_list(pool_3d_ops_list,
pool_3d_configs_short + pool_3d_configs_long,
Pool3dBenchmark)
if __name__ == "__main__":
op_bench.benchmark_runner.main()
|
